Method and system for stabilizing wire arcs of IC chip and memory chip of wire bonding machine

Abstract

The invention discloses a wire arc stabilizing method and system for an IC chip and a memory chip of a wire bonding machine, and belongs to the technical field of semiconductor packaging equipment control. The method comprises the following steps: extracting bonding wire process parameters, and generating initial path data containing key points; performing smooth fitting on the initial path data by adopting a high-order parameterized curve to generate a three-dimensional geometric path with curvature continuity; constructing a multi-order kinematic model, and coupling the geometric path with the physical system dynamics performance; generating a constraint curve distributed along the path based on curvature characteristics of the geometric path and dynamic constraints of dynamics; generating a smooth motion curve through a planning algorithm based on the constraint curve; synthesizing the motion curve and the geometric path, generating a feed-forward motion control instruction, transmitting the feed-forward motion control instruction to a digital signal processor, and transmitting the feed-forward motion control instruction to a driver for execution; through a hierarchical control strategy for decoupling geometric path planning and motion speed planning, high-speed, high-precision and high-stability 3D line arc trajectory planning is realized.

Description

Technical Field

[0001] This application belongs to the field of semiconductor packaging equipment control technology, specifically relating to a method and system for stabilizing the arc of IC chips and memory chips in a wire bonding machine. Background Technology

[0002] With the evolution of semiconductor technology, integrated circuit chips are developing towards miniaturization, high density, and multifunctionality. Whether it's high-end memory chips, logic chips, or complex system-in-package (SoC) packages, their internal interconnect structures are becoming increasingly complex. Wire bonding, as the core process for chip electrical interconnection, directly determines the yield and efficiency of packaging due to its motion control performance.

[0003] In high-end IC packaging, to accommodate thinner packages and more complex wiring requirements, ultra-low arc lengths, long-span arc lengths, and multi-layer 3D arc lengths are becoming increasingly common. This poses a significant challenge to the motion trajectory planning of wire bonding machines.

[0004] When faced with complex processes such as ultra-low arc length (ULL), long-span arc length, and multi-layer three-dimensional arc length widely used in high-end IC packaging, traditional trajectory planning algorithms often employ simple velocity planning (5th order). These methods only focus on the start and end points, often lacking in-depth optimization of three-dimensional spatial motion coupling. Furthermore, they struggle to perfectly suppress residual mechanical vibrations caused by sudden jerk changes across the entire velocity range, leading to arc instability and product defects. In addition, traditional linear interpolation motion during high-speed wire drawing can easily cause continuous unidirectional rigid friction between the wire bonding material (gold wire) and the inner wall of the bonding tool (ceramic nozzle), resulting in fluctuations in wire release resistance and consequently causing process defects such as unstable arc formation or even wire breakage. While these minute vibrations and unstable wire resistance might be negligible in ordinary processes, in high-end applications where space is extremely limited or wire rigidity is poor, they can easily cause wire collapse, misalignment, or even neck breakage. This leads to poor consistency in wire arc formation, failing to meet the stringent requirements of modern packaging processes for tracking dynamic errors in long-distance, high-speed wire arc movements (<5µm, 7mm wire length, XY acceleration 30G). To address this issue, existing solutions typically resort to reducing operating speed or increasing settling time to minimize, but not completely eliminate, defect rates. This not only makes it difficult for equipment to balance high speed and high precision but also significantly reduces production efficiency, failing to meet the semiconductor packaging and testing industry's dual demands for high output and high quality.

[0005] Therefore, developing a general trajectory planning method based on high-order dynamic constraints to meet the packaging needs of various high-end chips is of great significance for improving the overall competitiveness of semiconductor packaging equipment. Summary of the Invention

[0006] Aiming to address the problems in existing trajectory planning algorithms, such as poor line arc consistency due to strong coupling between speed and geometry, high-speed residual vibration caused by lack of high-order dynamic constraints, spatial interference risk under high-density multilayer packaging, and wire release instability caused by linear motion friction, this invention provides a method and system for stabilizing the line arc of IC chips and memory chips in wire bonding machines. By adopting a decoupling strategy between geometric path planning and motion speed planning, it achieves high-speed, high-precision, and high-stability wire bonding trajectory planning, effectively suppressing vibration and interference, and improving the quality of line arc forming and production efficiency.

[0007] The technical solution is as follows: On the one hand, a method for stabilizing the arc of IC chips and memory chips in wire bonding machines is provided, which adopts a decoupling strategy of geometric path planning and motion speed planning, including: Extract wire bonding process parameters and generate initial path data containing key points; The initial path data is smoothly fitted using a parametric curve to generate a three-dimensional geometric path with curvature continuity; A multi-order dynamic model is constructed, the three-dimensional geometric path is coupled with the physical time domain and subjected to arc parameterization, and a mapping relationship between the geometric space domain and the physical time domain is established. Based on the curvature characteristics of the geometric path and the limits of the mechanics, actuators, and motion controllers, dynamic constraints are formed to generate smooth constraint curves of position, velocity, acceleration, and jerk distributed along the path. Based on the constraint curve, a smooth position, velocity, acceleration, and jerk motion curve is generated by a planning algorithm. The motion curve and the three-dimensional geometric path are synthesized in full state. Through the multi-order dynamic model, a feedforward control command containing full state information of position, velocity, acceleration and jerk is generated and transmitted to the digital signal processor (DSP) and then sent to the driver for execution.

[0008] Preferably, the parameterized curve is a high-order Bézier curve, a B-spline curve, or a NURBS curve.

[0009] Preferably, the three-dimensional geometric path is a three-dimensional spatial surface path with helical features or composite curvature.

[0010] Preferably, the dynamic constraints include multiple independent constraint sub-models, which include one or more of the following: inherent performance constraints of the mechanism, path dynamics interaction constraints, higher-order smoothness constraints, and trajectory accuracy and process constraints.

[0011] Preferably, the dynamic constraints are a set of theoretical upper limits derived from various constraints.

[0012] Preferably, the planning algorithm is a multi-stage velocity planning algorithm with acceleration continuity, which identifies the low-speed zone and plans a smooth acceleration and deceleration process by performing global bidirectional scanning and look-ahead processing on the constraint curve.

[0013] Preferably, the multi-stage velocity planning algorithm for jerk continuity is a seven-segment S-shaped interpolation algorithm.

[0014] Furthermore, the method also includes a parallel computing optimization step, which optimizes the underlying computation instructions and performs multi-threaded concurrent processing on the fitting algorithm for the parameterized curve.

[0015] On the other hand, a system for stabilizing the arc of IC chips and memory chips in wire bonding machines is provided, which is used to implement the above-mentioned method for stabilizing the arc of IC chips and memory chips in wire bonding machines. The system includes: The process parameter parsing module is used to receive and parse wire bonding process parameters and generate critical path points; The geometric path fitting module is connected to the process parameter analysis module and is used to smoothly fit the key path points using parametric curves to generate a three-dimensional geometric path. The motion planning module, connected to the geometric path fitting module, is used to form dynamic constraints based on the curvature characteristics of the three-dimensional geometric path and the limits of the mechanics, actuators, and motion controllers, and to generate position, velocity, acceleration, and jerk constraint curves distributed along the path. The motion planning module includes at least a kinematic modeling and mapping unit, a constraint analysis and envelope generation unit, and a velocity planning unit. The trajectory synthesis module is connected to the geometric path fitting module and the motion planning module respectively. It is used to synthesize full-state information including position, velocity, acceleration and jerk, generate feedforward control commands and transmit them to the digital signal processor (DSP), and then send them down to the driver for execution.

[0016] Furthermore, it also includes a parallel optimization module for improving the computational processing speed of the geometric path fitting module and / or the motion planning module.

[0017] Compared with existing technologies, this technical solution has at least the following technical advantages: This application employs a decoupled, hierarchical control strategy through geometric path planning and motion velocity planning. First, it utilizes high-order parametric curve fitting to generate a three-dimensional geometric path with a fixed shape and continuous curvature. Combined with its convex hull property, this provides absolute anti-interference safety for high-density wiring, and the wire laying friction is optimized through three-dimensional surface design. Then, based on a PVAJ multi-order dynamic model incorporating jerk and a multi-dimensional constraint architecture, a maximum allowable velocity curve is generated. A jerk-continuous planning algorithm with global look-ahead capability is used to achieve time-optimal smooth motion control while ensuring micron-level tracking accuracy. This method eliminates trajectory drift, residual vibration, and wire laying instability caused by velocity-shape coupling and abrupt acceleration changes at the source. Thus, under high-speed operating conditions, it simultaneously achieves high consistency, high precision, high stability, and high production efficiency in wire arc forming, providing a universal solution for high-end chip packaging.

[0018] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] Figure 1 The waveform diagram of the actual ultra-low arc after using the wire bonding machine IC chip and memory chip to stabilize the arc according to a preferred embodiment of this application; Figure 2 The waveform diagram of the actual CSP wire arc after using the wire bonding machine IC chip and memory chip to stabilize the wire arc according to a preferred embodiment of this application; Figure 3 A flowchart of a method for stabilizing wire arcs in an IC chip and a memory chip for a wire bonding machine, provided as a preferred embodiment of this application; Figure 4 This application provides a preferred embodiment of a system structure diagram for stabilizing wire arcs using an IC chip and memory chip in a wire bonding machine. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] Explanation of related terms: Position (P): The coordinates of an object in space.

[0024] Velocity (V): The rate of change of position with time, i.e., the first derivative of displacement.

[0025] Acceleration (A): The rate of change of velocity with time, i.e., the first derivative of velocity and the second derivative of position.

[0026] Jerk (J): The rate of change of acceleration with time, i.e., the first derivative of acceleration, the second derivative of velocity, and the third derivative of position.

[0027] Example 1 like Figure 3 As shown, a method for stabilizing wire arcs in IC chips and memory chips for wire bonding machines is provided, comprising the following steps: Phase 1: Geometric path planning. The goal of this phase is to generate a smooth, absolutely definite three-dimensional geometric path that is independent of the speed of motion.

[0028] Step S101, wire bonding process parameter analysis and critical path point generation: Analyze the wire bonding process parameters and generate initial path data containing critical path points for different levels of wire arcs.

[0029] First, the system receives preliminary wire bonding path data output from the upper-level process design module. This data is generated based on preset first and second solder joints and related process parameters. The system then parses these process parameters to generate a set of critical path points defining the basic shape of the wire arc, and outputs a discrete sequence of critical path point coordinates.

[0030] The wire bonding path data physically constitutes the geometric skeleton features that define the shape of the wire bonding trajectory, including at least the coordinates of the first weld point, the second weld point, and the highest point of the arc. To achieve finer shape control, the wire bonding path data may further include auxiliary control points, such as reverse bend points and trajectory transition points, used to adjust the geometric tension of the arc and enhance the stability of the weld point neck. These key path points collectively define the topological structure of the wire bonding trajectory, providing fundamental geometric constraints for subsequent higher-order fitting.

[0031] Step S102, Geometric path construction based on high-order fitting: The wire bonding path data (key path point coordinate sequence) is smoothly fitted using parametric curves to generate a three-dimensional geometric path with curvature continuity.

[0032] An initial path, which may contain bends, is generated based on the critical path points, and more feature points are sampled according to a strategy. The feature points are then smoothly fitted using a high-order parametric curve, and finally a parametric, smooth 3D geometric path is output.

[0033] After obtaining the wire bonding path data (i.e. the critical path point coordinate sequence) in step S101, the system performs the transformation from discrete process parameters to a continuous high-order smooth trajectory, which specifically includes initial path modeling, feature extraction, and smooth fitting based on a high-order parameterized model.

[0034] First, for each independent geometric span, the system generates an initial 3D path through mathematical modeling and extracts several discrete feature points as fitting benchmarks according to a preset sampling strategy. Then, the system selects a parametric curve model with convex hull properties and higher-order derivative continuity to approximate and smooth the span marked by the discrete feature points. The core of this process lies in using the mathematical properties of the parametric curve to eliminate geometric bends in the initial path, generating a final geometric trajectory with curvature continuity.

[0035] In specific embodiments, the parametric curve model is preferably a high-order Bézier curve, but B-spline curves or NURBS curves can also be used.

[0036] In one embodiment, a higher-order Bézier curve is used. Taking a fifth-order Bézier curve as an example, its mathematical expression can be described as follows: in, The order of the curve. For the first Control point vectors for The Bernstein basis function is defined as follows: in, The coefficients are binomial coefficients.

[0037] In this mode, the generated trajectory Attracted by the weights of each control point, an inscribed approximation is formed within the control polygon. This characteristic establishes a defined spatial safety boundary in engineering applications: as long as the control points are set to prevent interlayer interference, the actual smooth curve generated based on the convex hull property will naturally converge within this safety region. This effectively avoids the trajectory overshoot or path expansion phenomenon caused by high-speed motion in traditional interpolation algorithms, providing a reliable geometric safety margin for high-density multilayer encapsulation.

[0038] Furthermore, to address the rigid friction problem between the welding tool and the wire in traditional linear motion, this invention constructs a three-dimensional geometric path generated by a parametric curve. This three-dimensional geometric path is a three-dimensional spatial surface path with helical features or composite curvature. This three-dimensional flexible trajectory allows the welding head to guide the release of the welding wire material with a gradually changing entry angle during movement, thereby transforming the unidirectional rigid friction in traditional linear motion into dispersed flexible contact. This geometric characteristic effectively reduces the dynamic frictional resistance between the welding wire and the inner wall of the tool, avoiding the problem of unstable wire release, thus significantly improving the geometric stability and yield of the final formed arc.

[0039] Furthermore, the parametric curve model is a high-order fitting algorithm with adaptive optimization capabilities for asymmetric paths. When there is a significant length difference between adjacent trajectory segments, the generated smooth curve can automatically adjust the rate of change of the tangential angle, enabling the welding head to quickly establish normal acceleration on the short side to escape the interference zone, while completing tangential contact earlier on the long side to utilize the straight line segment for acceleration, thereby achieving time-optimal motion allocation on the micro-path.

[0040] To meet the real-time requirements of high-speed production lines for the aforementioned intensive matrix operations and polynomial interpolation, this method also includes a parallel computing optimization step. By optimizing the underlying computational instructions and implementing multi-threaded concurrent processing for the core high-order fitting algorithm, the system can complete the entire process path calculation within an extremely short time period (<1ms), ensuring the real-time performance of trajectory planning and the continuous high-speed operation of the equipment.

[0041] Step S103: Construct kinematic model and parameter mapping: Couple the three-dimensional geometric path with the physical time domain and perform arc parameterization to establish the mapping relationship between the geometric space domain and the physical time domain.

[0042] In this step, the constructed kinematic model is a multi-dimensional PVAJ (position-velocity-acceleration-jerk) full-state kinematic model, which deeply couples the three-dimensional geometric path generated in step S102 with the dynamic performance of the physical system, and establishes a precise mapping relationship between the geometric spatial domain and the physical time domain.

[0043] Step S1031: Obtain the physical constraint limit.

[0044] First, the system acquires the global dynamic constraint boundaries preset for specific actuators (such as servo motors and robotic arms). These constraint boundaries constitute the feasible domain of motion planning, mainly including but not limited to the system's maximum operating speed, the maximum allowable acceleration to prevent mechanical overload (especially centripetal acceleration in curvilinear motion), the maximum allowable jerk to suppress residual vibration, and the maximum allowable dynamic path deviation to ensure trajectory tracking accuracy.

[0045] Step S1032: Construct a 3D PVAJ (position-velocity-acceleration-jerk) motion model.

[0046] Based on this, to ensure that the final generated motion commands are strictly controlled within the aforementioned physical limits in the four dimensions of position, velocity, acceleration, and jerk, the system constructs a 3D PVAJ motion model containing higher-order derivative terms. This invention establishes a kinematic state equation connecting the geometric curve parameter domain and the physical time domain based on the chain rule for finding the derivative of composite functions.

[0047] For any time The system will locate Defined as a parameterized curve The output of is then used to derive the following: Location: ; speed: ; Acceleration: ; accelerometer: ; in, For the geometric parameters of the curve, , , These are its first, second, and third derivatives with respect to time, respectively; , , These represent the geometric path with respect to the parameters. The first to third geometric derivatives. This set of equations reveals that the physical state of motion depends not only on the curvature characteristics of the geometric path itself, but also on the dynamic laws governing the evolution of parameters over time.

[0048] Step S1033: Arc length parameterization.

[0049] Furthermore, due to the parameters of the parametric curve (taking Bézier curves as an example) Physical length of geometric curve It exhibits a non-linear relationship, directly affecting the parameters. Interpolation can cause fluctuations in the actual linear velocity.

[0050] In one embodiment, an arc length parameterization mechanism is introduced. This mechanism establishes parameters... Cumulative physical arc length of the curve Monotonic mapping function between This eliminates the influence of uneven geometric parameter distribution on velocity planning. Its mapping relationship can be characterized as the integral of the geometric derivative magnitude over the parameter domain: in, This indicates the curve from the starting point to the current parameter. The cumulative physical arc length at the location; For integration variables; , , Let represent the first derivatives of the coordinates with respect to the parameters. Integrand Indicates the curve at The instantaneous geometric rate at a given moment.

[0051] Through the above mapping, the tangential vector magnitude at any point on the curve can be accurately calculated, thus enabling the calculation of the magnitude based on the physical arc length. The planned velocity profile is precisely inversely solved back to the geometric parameter domain. This ensures that when the actuator is running on a complex three-dimensional curve, its actual synthesized speed strictly follows the preset dynamic expectations, avoiding unexpected acceleration and deceleration oscillations caused by parameter nonlinearity.

[0052] Step S104, Generate constraint curves based on physical limits: Based on the curvature characteristics of the geometric path and the limits of the mechanism, actuator, and motion controller, form dynamic constraints and generate smooth constraint curves of position, velocity, acceleration, and jerk (jump) distributed along the path.

[0053] Before generating the time-domain trajectory, the system first automatically calculates a multi-dimensional constraint set distributed along the entire path based on the curvature characteristics of the geometric path. This set represents the theoretical upper limit that can be achieved at each geometric location point on the path, constrained by the physical performance and manufacturing precision of the actuator.

[0054] This invention proposes an open multidimensional constraint architecture, in which the final effective motion state envelope is determined by the minimum set of multiple independent constraint sub-models.

[0055] The constraint sub-model includes, but is not limited to, one or more of the following types of physical constraints: Step S1041: Inherent performance constraint class of the mechanism.

[0056] These constraints are based on the physical limits of the actuator itself and are designed to prevent hardware damage or exceeding the operating range. They include, but are not limited to: Absolute speed limit: The basic speed limit is determined by the maximum speed characteristics of the servo motor, the upper limit of the encoder sampling frequency, or the maximum allowable linear speed of the mechanical guide rail. And the global acceleration limit determined by the maximum current of the driver. and accelerometer limit .

[0057] Torque and overload limits: Speed ​​or acceleration boundaries set based on the motor's peak torque or the driver's maximum current.

[0058] Step S1042: Path dynamics interaction constraint class.

[0059] These constraints arise from the dynamic interaction between the moving body and the bending path, and are designed to prevent dynamic instability. They include, but are not limited to: Centripetal acceleration (normal force) constraint: When the actuator passes through a path segment with a certain curvature, in order to prevent motor torque overload, lateral deformation of the mechanical structure, or end effector vibration caused by excessive normal acceleration, the system uses the following formula: in, For path curvature.

[0060] The upper limit of speed derived by reverse calculation The corresponding upper limit of dynamic tangential acceleration .

[0061] Tangential dynamic constraints: Dynamic restrictions on tangential acceleration for a specific direction of gravity or distribution of friction.

[0062] Step S1043: Higher-order smoothness constraint class Such constraints aim to suppress higher-order vibrations and improve motion stability. These include, but are not limited to: Geometric jerk constraint: To suppress flexible impacts caused by drastic changes in geometry (such as abrupt curvature changes), this invention is based on the coupling relationship between jerk and higher-order geometric derivatives in the kinematic model. The derivation shows that it is limited by the maximum permissible jerk. Speed ​​limit and dynamic jerk constraints .

[0063] Step S1044, Trajectory Accuracy and Process Constraints.

[0064] These constraints are based on the final process quality indicators. They include, but are not limited to: Dynamic chord height error constraint: To ensure that the maximum deviation between the actual trajectory and the theoretical curve during discretization interpolation does not exceed a preset threshold. The system combines the sampling period of the control system and back-calculates the upper limit of speed based on the principles of circular arc approximation or Taylor expansion. And acceleration limits.

[0065] Process-specific speed limits: Pre-set mandatory low-speed limits for specific sensitive areas in the welding process (such as the heat-affected zone or reverse bending points).

[0066] By combining the various constraint sub-models described above, the system calculates the value of any point on the path. The final multidimensional motion constraint envelope at the location : Among them, velocity envelope And at any position The actual acceleration and acceleration command at the point are both restricted to the dynamic range after dynamic decoupling. and Within the range.

[0067] Step S105: Continuous velocity planning based on global look-ahead: Based on the constraint curve, a smooth position, velocity, acceleration and jerk motion curve is generated by a planning algorithm.

[0068] After establishing the maximum permissible speed envelope Based on dynamic acceleration and jerk constraints, the system employs a multi-stage velocity planning algorithm with continuous jerk. In a preferred embodiment, a seven-segment S-curve interpolation algorithm can be used to plan a smooth and time-optimal actual velocity curve. .

[0069] Specifically, the system along the path Perform a global bidirectional scan and look-ahead processing. This process aims to identify "low-speed bottlenecks" on the path: when a low-speed section is detected ahead, limited by curvature or dynamic accuracy, the system determines the maximum allowable dynamic deceleration based on the current path point. The deceleration starting point is calculated in reverse to ensure that the actual velocity of the moving object has smoothly decreased to below a safe threshold when it reaches the bottleneck position. Simultaneously, to ensure the continuity of the rate of change of acceleration, the velocity planning model constructed in this invention strictly follows the rule of "constant or continuously changing jerk," satisfying... Given the constraints, a complete velocity profile is generated that includes acceleration, constant velocity, and deceleration segments.

[0070] The final planning output is represented as a function of the physical arc length with respect to time. and the corresponding physical speed .

[0071] Step S106, High-order trajectory synthesis and feedforward control command generation: The motion curve and the three-dimensional geometric path are synthesized in full state. Through the multi-order dynamic model, feedforward control commands containing full state information of position, velocity, acceleration and jerk are generated and transmitted to the digital signal processor (DSP), and then sent to the driver for execution.

[0072] In this step, the system will use the time-domain velocity curve planned in step S105. The geometric space curve constructed in step S102 Perform full-state synthesis. Using the kinematic mapping model established in step S1032, calculate the time of each control cycle. Precise motion state quantities, including position ,speed acceleration And accelerometer .

[0073] Through the above synthesis steps, feedforward control commands containing high-order dynamic information can be output to the underlying servo control loop. This full-state feedforward mechanism enables the wire bonder to adjust the torque output in advance based on predicted dynamic requirements when executing micron-level high-speed complex trajectories, thereby achieving near-zero phase lag precise tracking and eliminating mechanical resonance caused by sudden command changes at the source, significantly improving the process stability of high-speed welding.

[0074] In another embodiment, a system for stabilizing arcs in IC chips and memory chips of a wire bonding machine is provided, which is used in the method for stabilizing arcs in IC chips and memory chips of a wire bonding machine provided in Embodiment 1, including the following modules (such as... Figure 4 (as shown) The process parameter parsing module is used to receive and parse wire bonding process parameters and generate critical path points; The geometric path fitting module is connected to the process parameter analysis module and is used to smoothly fit the key path points using parametric curves to generate a three-dimensional geometric path. The motion planning module, connected to the geometric path fitting module, is used to generate dynamic constraints based on the curvature characteristics of the three-dimensional geometric path and the limits of the mechanics, actuators, and motion controllers, and to generate position, velocity, acceleration, and jerk constraint curves distributed along the path. The motion planning module includes at least a kinematic modeling and mapping unit, a constraint analysis and envelope generation unit, and a velocity planning unit.

[0075] The kinematic modeling and mapping unit is used to establish the kinematic mapping relationship between the three-dimensional geometric path and the physical time domain; The constraint analysis and envelope generation unit is used to generate a set of multidimensional constraints distributed along the three-dimensional geometric path based on the multidimensional physical constraints. The velocity planning unit is used to plan a time-domain velocity profile with continuous acceleration under the constraints of the multidimensional motion constraint envelope.

[0076] The trajectory synthesis module is connected to the geometric path fitting module and the motion planning module respectively. It is used to synthesize full-state information including position, velocity, acceleration and jerk, generate feedforward control commands and transmit them to the digital signal processor (DSP), and then send them down to the driver for execution.

[0077] In another embodiment, it further includes a parallel optimization module for improving the computational processing speed of the geometric path fitting module and / or the motion planning module.

[0078] from Figure 1 As can be seen from the data, the smooth characteristics of the actual trajectory curve indicate that the jerk generated by the method of stabilizing the arc of the IC chip and memory chip in the wire bonding machine is continuous, which effectively suppresses the residual vibration of the robotic arm during high-speed movement.

[0079] from Figure 2 As can be seen, the actual waveform closely matches the ideal CSP arc shape, indicating that the method for stabilizing the arc of IC chips and memory chips in the wire bonding machine has extremely high servo tracking performance, minimal hysteresis error, and can accurately reproduce the preset process parameters.

[0080] contrast Figure 1 and Figure 2 It is evident that the method for stabilizing the arc of IC chips and memory chips in this wire bonding machine can generate high-quality motion curves for different arc types. This demonstrates that the technical solution of this application has good versatility, can adapt to the diverse requirements of different packaging forms for arc geometry, and can maintain motion stability under different height spans.

[0081] In summary, the present invention aims to provide a method and system for stabilizing the arc of IC chips and memory chips in wire bonding machines. This method and system is based on high-order curve fitting, three-dimensional spatial curves, and multi-order dynamic constraints to stabilize the arc of IC chips and memory chips in wire bonding machines.

[0082] Specifically, this invention employs a layered control strategy that decouples geometric path planning and motion velocity planning. First, at the geometric path planning layer, the system generates initial three-dimensional spatial paths that meet process requirements for different levels of line arcs through multi-layer process parameter analysis. Then, using this initial path as a reference, high-precision smooth fitting is performed using high-order parametric curves with spatial boundary constraints. This step utilizes the mathematical properties of high-order curves to ensure the fitting path highly overlaps with the initial modeling path while maintaining the curvature continuity of the entire path. It also establishes a defined anti-interference safety boundary using the convex hull properties of curve control points, and constructs flexible three-dimensional contact features to optimize line-laying friction. Based on the determined geometric path, the system enters the motion velocity planning layer, introducing a multi-order dynamic model including position (P), velocity (V), acceleration (A), and jerk (J). Through a velocity planning algorithm based on dynamic constraints, a smooth, continuous velocity profile is decoupled and independently planned while satisfying the maximum dynamic deviation and higher-order derivative limitations of the mechanical system, ultimately synthesizing the full-state motion command.

[0083] The hierarchical control strategy that decouples geometric path planning and motion velocity planning has the following technical advantages: 1. Ensure high consistency of arc shape: Ensure that the actual generated weld wire trajectory remains constant in geometric space regardless of the production capacity (UPH) mode or speed level of the equipment, completely eliminating the arc radius drift or path deformation caused by speed adjustment in traditional algorithms. It is especially suitable for ultra-low arc (ULL) and long arc applications with extremely high requirements for line shape accuracy.

[0084] 2. Achieve ultra-smooth motion at high speeds: While ensuring path tracking accuracy, continuous jerk control throughout the entire process eliminates the flexible impact caused by sudden acceleration changes at the source, significantly reducing mechanical residual vibration during high-speed start-up and shutdown and three-dimensional turning of the wire bonding machine, thereby improving the yield of micro-pitch welding.

[0085] 3. Maximize overall production efficiency: While meeting the maximum dynamic error limit, fully explore the physical performance limits of servo motors and mechanical structures to achieve optimal motion planning in time, thereby significantly improving the welding output of the equipment while ensuring forming quality.

[0086] 4. Solving the Spatial Interference Challenge of High-Density Multilayer Wire Bonding: For increasingly complex multilayer stacked chip packaging and high-density multilayer wire bonding processes, this invention can pre-set precise geometric parameters and solidify the path for the specific height and span requirements of each wire layer. Because the trajectory shape of this invention does not drift with speed changes, it ensures that adjacent gold wires do not touch each other in extremely small spatial intervals, effectively preventing short-circuit risks caused by high-speed trajectory overflow, and meeting the stringent requirements of chip miniaturization for extreme space utilization.

[0087] 5. Improved stability and smoothness of wire arc forming: By constructing a three-dimensional curved surface path, this invention changes the contact mode between the welding material and the tool's inner wall in traditional linear motion, transforming rigid friction into flexible contact. This not only reduces wire feeding resistance but also effectively avoids wire tearing or unstable forming caused by friction fluctuations, significantly improving the geometric quality and long-term process reliability of the wire arc.

[0088] This invention achieves a layered control strategy that decouples geometric path planning and motion speed planning, enabling the maintenance of micron-level trajectory repeatability and positioning accuracy and effective suppression of device resonance under arbitrary speed settings. It also provides absolute spatial safety margin for multi-layer complex packaging and significantly improves wire laying quality. Thus, it provides a universal solution for packaging various high-end integrated circuit chips that takes into account high speed, high precision, high safety and high process quality.

[0089] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

Claims

1. A method for stabilizing arc of a wire bonder IC chip and memory chip, characterized in that, Adopting a decoupling hierarchical control strategy of geometric path planning and motion velocity planning, comprising: extracting wire bonding process parameters, generating initial path data containing key points; adopting a parametric curve to smooth fit the initial path data, generating a three-dimensional geometric path with continuous curvature; constructing a multi-order dynamic model, coupling the three-dimensional geometric path with the physical time domain and performing line-arc parameterization processing, establishing a mapping relationship between the geometric spatial domain and the physical time domain; forming dynamic dynamic constraints based on the curvature characteristics of the geometric path and the limits of the machine, driver, and motion controller, generating position, velocity, acceleration, and jerk smooth constraint curves distributed along the path; based on the constraint curve, generating smooth position, velocity, acceleration, and jerk motion curves through a planning algorithm; synthesizing the motion curve with the three-dimensional geometric path in full state, generating a feedforward control instruction containing position, velocity, acceleration, and jerk full state information through the multi-order dynamic model and transmitting it to the digital signal processor, and then issuing it to the driver for execution.

2. The method of claim 1, wherein the wire bonding machine is a thermal compression bonding machine. The parametric curve adopts a high-order Bezier curve, a B-spline curve, or a NURBS curve.

3. The method of claim 1, wherein the wire bonding machine IC chip and memory chip stabilizing arc comprises: a wire bonding machine IC chip; a memory chip; a wire bonding machine; a wire bonding machine IC chip and memory chip stabilizing arc; and a wire bonding machine IC chip and memory chip stabilizing arc. The three-dimensional geometric path is a three-dimensional spatial curved surface path with spiral characteristics or composite curvature.

4. The method of claim 1, wherein the wire bonding machine IC chip and memory chip stabilizing arc comprises: a wire bonding machine IC chip; a memory chip; a wire bonding machine; a wire bonding machine IC chip and memory chip stabilizing arc; and a wire bonding machine IC chip and memory chip stabilizing arc. The dynamic dynamic constraints include multiple independent constraint sub-models, which include one or more of mechanism inherent performance constraints, path dynamic interaction constraints, high-order smoothness constraints, and trajectory accuracy and process constraints.

5. The method of claim 4, wherein the wire bonding machine IC chip and memory chip stabilizing arc comprises: a wire bonding machine IC chip and memory chip stabilizing arc. The dynamic dynamic constraints are a set of theoretical upper limits derived from various constraints.

6. The method of claim 1, wherein the wire bonding machine IC chip and memory chip stabilizing arc comprises: a wire bonding machine IC chip; a memory chip; a wire bonding machine; a wire bonding machine IC chip and memory chip stabilizing arc; and a wire bonding machine IC chip and memory chip stabilizing arc. The planning algorithm is a jerk continuity multi-stage velocity planning algorithm that identifies low-speed areas and plans smooth acceleration and deceleration processes through global bidirectional scanning and forward-looking processing of the constraint curve.

7. The method of claim 6, wherein the wire bonding machine IC chip and memory chip stabilizing arc comprises: a wire bonding machine IC chip and memory chip stabilizing arc. The jerk continuity multi-stage velocity planning algorithm is a seven-segment S-type interpolation algorithm.

8. The method of claim 1, wherein the wire bonding machine IC chip and memory chip stabilizing arc comprises: a wire bonding machine IC chip; a memory chip; a wire bonding machine; a wire bonding machine IC chip and memory chip stabilizing arc; and a wire bonding machine IC chip and memory chip stabilizing arc. The method further includes a parallel computing optimization step for optimizing the fitting algorithm of the parametric curve in terms of bottom-level operation instructions and multi-thread concurrent processing.

9. A system for stabilizing a wire arc of an IC chip and a memory chip by a wire bonder for implementing the method for stabilizing a wire arc of an IC chip and a memory chip according to any one of claims 1 to 8, characterized by Comprising: a process parameter analysis module for receiving and analyzing wire bonding process parameters to generate key path points; a geometric path fitting module connected to the process parameter analysis module for smoothing fitting the key path points using a parametric curve to generate a three-dimensional geometric path; a motion planning module connected to the geometric path fitting module for forming dynamic dynamic constraints based on the curvature characteristics of the three-dimensional geometric path and the limits of the machine, driver, and motion controller, generating position, velocity, acceleration, and jerk constraint curves distributed along the path, the motion planning module at least including a kinematics modeling and mapping unit, a constraint analysis and envelope generation unit, and a velocity planning unit; a trajectory synthesis module connected to the geometric path fitting module and the motion planning module, respectively, for synthesizing position, velocity, acceleration, and jerk full state information to generate a feedforward control instruction and transmit it to a digital signal processor, and then issue it to the driver for execution.

10. The system for stabilizing the arc of a wire bonder IC chip and memory chip according to claim 9, wherein, It also includes a parallel optimization module for improving the operation processing speed of the geometric path fitting module and / or the motion planning module.

Images